Chronic pain affects over 100 million Americans, and arthritis is the most common cause. Existing treatments for chronic arthritic pain are only mildly effective, and risks of medications used to treat pain are numerous and continue to be discovered. Treatment of chronic is a high priority research area for VA CSR&D. Naltrexone is an opioid antagonist that is FDA approved in an oral daily dose of 50 mg to prevent recidivism in alcoholics. At much lower doses of 4 ? 4.5 mg daily, however, it has been shown in small, blinded, randomized trials to improve pain in fibromyalgia, gastrointestinal symptoms in Crohn?s disease, and quality of life in multiple sclerosis. The only other published data are case reports in complex regional pain syndrome, low back pain, and scleroderma. However, advocacy of low-dose naltrexone (LDN) by internet-based MDs and patients is high, and since LDN can be prescribed off-label, its use greatly exceeds what is justified by evidence. The drug can be prescribed only via compounding pharmacies, so its use costs a patient ~$40/month. Among the many unproven treatments that are widely used, LDN is of particular interest because results of surveys of patients are particularly impressive, because it is quite safe, and because its benefit is plausible pharmacologically. There is evidence both for modulation of central pain-processing pathways and for down- regulation of inflammatory pathways in microglia. Considering the diversity of conditions proposed to benefit from LDN and the unequivocal need for better approaches to pain relief in chronic conditions, high-quality clinical trials are needed in both inflammatory and non-inflammatory conditions. This small but placebo- controlled study, powered to detect an effect size as small as that seen with NSAIDs or the most beneficial non-pharmacologic approaches, is proposed as a prerequisite for considering a pivotal trial through the VA Cooperative Studies Program. The proposed study is a randomized, double-blinded, cross-over, placebo-controlled trial in adults with osteoarthritis or inflammatory arthritis and persistent pain. Sixty patients will be enrolled for 12 weeks, during which they will receive LDN for 8 weeks and placebo for 4 weeks. Widely accepted patient-reported outcome measures will be used. The co-primary endpoints are reduction in pain severity or pain?s interference with function during 8 weeks of LDN compared to 4 weeks placebo, using the Brief Pain Inventory. Other patient- reported data will be used both as secondary outcomes and as covariates in analyzing determinants of response to treatment. Key inclusion criteria include diagnosis of OA or IA and pain rated at least 4 on a scale of 0-10, a widely accepted criterion. Key exclusion criteria, chosen conservatively for the purposes of safety, include opioid use or severe liver, kidney, or psychiatric disease. OA, IA, and inadequate control of pain are sufficiently common that the study can be completed in two years at a single center.